there's a lot of misinformation about OpenNext and why it exists - going to lay it out here so i can reference it when people deploy nextjs outside of vercel, they expect something pretty simple since for most companies it's "just a frontend" they run into surprising issues with various features not working or behaving weirdly - you may not have seen these personally but they do exist historically there were fragmented efforts to reverse engineer these problems but none of them really stuck around given how much next changes and how complicated it has gotten OpenNext was created because people kept coming to us asking for help with this problem given we help people ship applications to AWS it primarily is the missing documentation about various nextjs internals, ideas on how to deploy certain complex features and their tradeoffs, quirks, workarounds, etc it also is a test suite that checks deployments to see if they are behaving correctly - these often break even on patch releases so it's quite tedious to keep the reverse engineering accurate finally it's some code that takes nextjs build output and massages it into something that can be deployed properly for some reason vercel employees keep saying that it is somehow a competitor to vercel trying to recreate vercel features that is not true - the focus is entirely on nextjs features listed in nextjs documentation. it does not bother with vercel features like feature flags, firewall, builds, domains, etc - no one is asking us for that it can't even create infrastructure for you - it serves as a collaborative effort so people deploying through terraform, sst, pulumi, k8s, even other hosting platforms can centralize efforts given how complicated nextjs has gotten, how coupled it is to vercel infra, and how intimately familiar we are with the hassle of self-hosting it, we often tell people to stay on vercel unless they have a good reason to move off (often it's price at scale) we really aren't out here actively focused on getting people to self-host their nextjs apps we deal with it because it's popular and we can't discriminate with our goal of helping people deploy anything even if it's the least important part of their system